4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2012 by Delphix. All rights reserved.
30 * ZAP - ZFS Attribute Processor
32 * The ZAP is a module which sits on top of the DMU (Data Management
33 * Unit) and implements a higher-level storage primitive using DMU
34 * objects. Its primary consumer is the ZPL (ZFS Posix Layer).
36 * A "zapobj" is a DMU object which the ZAP uses to stores attributes.
37 * Users should use only zap routines to access a zapobj - they should
38 * not access the DMU object directly using DMU routines.
40 * The attributes stored in a zapobj are name-value pairs. The name is
41 * a zero-terminated string of up to ZAP_MAXNAMELEN bytes (including
42 * terminating NULL). The value is an array of integers, which may be
43 * 1, 2, 4, or 8 bytes long. The total space used by the array (number
44 * of integers * integer length) can be up to ZAP_MAXVALUELEN bytes.
45 * Note that an 8-byte integer value can be used to store the location
46 * (object number) of another dmu object (which may be itself a zapobj).
47 * Note that you can use a zero-length attribute to store a single bit
48 * of information - the attribute is present or not.
50 * The ZAP routines are thread-safe. However, you must observe the
51 * DMU's restriction that a transaction may not be operated on
54 * Any of the routines that return an int may return an I/O error (EIO
58 * Implementation / Performance Notes:
60 * The ZAP is intended to operate most efficiently on attributes with
61 * short (49 bytes or less) names and single 8-byte values, for which
62 * the microzap will be used. The ZAP should be efficient enough so
63 * that the user does not need to cache these attributes.
65 * The ZAP's locking scheme makes its routines thread-safe. Operations
66 * on different zapobjs will be processed concurrently. Operations on
67 * the same zapobj which only read data will be processed concurrently.
68 * Operations on the same zapobj which modify data will be processed
69 * concurrently when there are many attributes in the zapobj (because
70 * the ZAP uses per-block locking - more than 128 * (number of cpus)
71 * small attributes will suffice).
75 * We're using zero-terminated byte strings (ie. ASCII or UTF-8 C
76 * strings) for the names of attributes, rather than a byte string
77 * bounded by an explicit length. If some day we want to support names
78 * in character sets which have embedded zeros (eg. UTF-16, UTF-32),
79 * we'll have to add routines for using length-bounded strings.
89 * The matchtype specifies which entry will be accessed.
90 * MT_EXACT: only find an exact match (non-normalized)
91 * MT_FIRST: find the "first" normalized (case and Unicode
92 * form) match; the designated "first" match will not change as long
93 * as the set of entries with this normalization doesn't change
94 * MT_BEST: if there is an exact match, find that, otherwise find the
95 * first normalized match
97 typedef enum matchtype
104 typedef enum zap_flags {
105 /* Use 64-bit hash value (serialized cursors will always use 64-bits) */
106 ZAP_FLAG_HASH64 = 1 << 0,
107 /* Key is binary, not string (zap_add_uint64() can be used) */
108 ZAP_FLAG_UINT64_KEY = 1 << 1,
110 * First word of key (which must be an array of uint64) is
111 * already randomly distributed.
113 ZAP_FLAG_PRE_HASHED_KEY = 1 << 2,
117 * Create a new zapobj with no attributes and return its object number.
118 * MT_EXACT will cause the zap object to only support MT_EXACT lookups,
119 * otherwise any matchtype can be used for lookups.
121 * normflags specifies what normalization will be done. values are:
122 * 0: no normalization (legacy on-disk format, supports MT_EXACT matching
124 * U8_TEXTPREP_TOLOWER: case normalization will be performed.
125 * MT_FIRST/MT_BEST matching will find entries that match without
126 * regard to case (eg. looking for "foo" can find an entry "Foo").
127 * Eventually, other flags will permit unicode normalization as well.
129 uint64_t zap_create(objset_t *ds, dmu_object_type_t ot,
130 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
131 uint64_t zap_create_norm(objset_t *ds, int normflags, dmu_object_type_t ot,
132 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
133 uint64_t zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
134 dmu_object_type_t ot, int leaf_blockshift, int indirect_blockshift,
135 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
136 uint64_t zap_create_link(objset_t *os, dmu_object_type_t ot,
137 uint64_t parent_obj, const char *name, dmu_tx_t *tx);
140 * Create a new zapobj with no attributes from the given (unallocated)
143 int zap_create_claim(objset_t *ds, uint64_t obj, dmu_object_type_t ot,
144 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
145 int zap_create_claim_norm(objset_t *ds, uint64_t obj,
146 int normflags, dmu_object_type_t ot,
147 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
150 * The zapobj passed in must be a valid ZAP object for all of the
151 * following routines.
155 * Destroy this zapobj and all its attributes.
157 * Frees the object number using dmu_object_free.
159 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
162 * Manipulate attributes.
164 * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
168 * Retrieve the contents of the attribute with the given name.
170 * If the requested attribute does not exist, the call will fail and
173 * If 'integer_size' is smaller than the attribute's integer size, the
174 * call will fail and return EINVAL.
176 * If 'integer_size' is equal to or larger than the attribute's integer
177 * size, the call will succeed and return 0. * When converting to a
178 * larger integer size, the integers will be treated as unsigned (ie. no
179 * sign-extension will be performed).
181 * 'num_integers' is the length (in integers) of 'buf'.
183 * If the attribute is longer than the buffer, as many integers as will
184 * fit will be transferred to 'buf'. If the entire attribute was not
185 * transferred, the call will return EOVERFLOW.
187 * If rn_len is nonzero, realname will be set to the name of the found
188 * entry (which may be different from the requested name if matchtype is
191 * If normalization_conflictp is not NULL, it will be set if there is
192 * another name with the same case/unicode normalized form.
194 int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name,
195 uint64_t integer_size, uint64_t num_integers, void *buf);
196 int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
197 uint64_t integer_size, uint64_t num_integers, void *buf,
198 matchtype_t mt, char *realname, int rn_len,
199 boolean_t *normalization_conflictp);
200 int zap_lookup_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
201 int key_numints, uint64_t integer_size, uint64_t num_integers, void *buf);
202 int zap_contains(objset_t *ds, uint64_t zapobj, const char *name);
203 int zap_prefetch_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
206 int zap_count_write(objset_t *os, uint64_t zapobj, const char *name,
207 int add, uint64_t *towrite, uint64_t *tooverwrite);
210 * Create an attribute with the given name and value.
212 * If an attribute with the given name already exists, the call will
213 * fail and return EEXIST.
215 int zap_add(objset_t *ds, uint64_t zapobj, const char *key,
216 int integer_size, uint64_t num_integers,
217 const void *val, dmu_tx_t *tx);
218 int zap_add_uint64(objset_t *ds, uint64_t zapobj, const uint64_t *key,
219 int key_numints, int integer_size, uint64_t num_integers,
220 const void *val, dmu_tx_t *tx);
223 * Set the attribute with the given name to the given value. If an
224 * attribute with the given name does not exist, it will be created. If
225 * an attribute with the given name already exists, the previous value
226 * will be overwritten. The integer_size may be different from the
227 * existing attribute's integer size, in which case the attribute's
228 * integer size will be updated to the new value.
230 int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
231 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
232 int zap_update_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
234 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
237 * Get the length (in integers) and the integer size of the specified
240 * If the requested attribute does not exist, the call will fail and
243 int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
244 uint64_t *integer_size, uint64_t *num_integers);
245 int zap_length_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
246 int key_numints, uint64_t *integer_size, uint64_t *num_integers);
249 * Remove the specified attribute.
251 * If the specified attribute does not exist, the call will fail and
254 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
255 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
256 matchtype_t mt, dmu_tx_t *tx);
257 int zap_remove_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
258 int key_numints, dmu_tx_t *tx);
261 * Returns (in *count) the number of attributes in the specified zap
264 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
267 * Returns (in name) the name of the entry whose (value & mask)
268 * (za_first_integer) is value, or ENOENT if not found. The string
269 * pointed to by name must be at least 256 bytes long. If mask==0, the
270 * match must be exact (ie, same as mask=-1ULL).
272 int zap_value_search(objset_t *os, uint64_t zapobj,
273 uint64_t value, uint64_t mask, char *name);
276 * Transfer all the entries from fromobj into intoobj. Only works on
277 * int_size=8 num_integers=1 values. Fails if there are any duplicated
280 int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx);
282 /* Same as zap_join, but set the values to 'value'. */
283 int zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj,
284 uint64_t value, dmu_tx_t *tx);
286 /* Same as zap_join, but add together any duplicated entries. */
287 int zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj,
291 * Manipulate entries where the name + value are the "same" (the name is
292 * a stringified version of the value).
294 int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
295 int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
296 int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value);
297 int zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
300 /* Here the key is an int and the value is a different int. */
301 int zap_add_int_key(objset_t *os, uint64_t obj,
302 uint64_t key, uint64_t value, dmu_tx_t *tx);
303 int zap_update_int_key(objset_t *os, uint64_t obj,
304 uint64_t key, uint64_t value, dmu_tx_t *tx);
305 int zap_lookup_int_key(objset_t *os, uint64_t obj,
306 uint64_t key, uint64_t *valuep);
308 int zap_increment(objset_t *os, uint64_t obj, const char *name, int64_t delta,
313 typedef struct zap_cursor {
314 /* This structure is opaque! */
317 struct zap_leaf *zc_leaf;
319 uint64_t zc_serialized;
325 int za_integer_length;
327 * za_normalization_conflict will be set if there are additional
328 * entries with this normalized form (eg, "foo" and "Foo").
330 boolean_t za_normalization_conflict;
331 uint64_t za_num_integers;
332 uint64_t za_first_integer; /* no sign extension for <8byte ints */
333 char za_name[MAXNAMELEN];
337 * The interface for listing all the attributes of a zapobj can be
338 * thought of as cursor moving down a list of the attributes one by
339 * one. The cookie returned by the zap_cursor_serialize routine is
340 * persistent across system calls (and across reboot, even).
344 * Initialize a zap cursor, pointing to the "first" attribute of the
345 * zapobj. You must _fini the cursor when you are done with it.
347 void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
348 void zap_cursor_fini(zap_cursor_t *zc);
351 * Get the attribute currently pointed to by the cursor. Returns
352 * ENOENT if at the end of the attributes.
354 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
357 * Advance the cursor to the next attribute.
359 void zap_cursor_advance(zap_cursor_t *zc);
362 * Get a persistent cookie pointing to the current position of the zap
363 * cursor. The low 4 bits in the cookie are always zero, and thus can
364 * be used as to differentiate a serialized cookie from a different type
365 * of value. The cookie will be less than 2^32 as long as there are
366 * fewer than 2^22 (4.2 million) entries in the zap object.
368 uint64_t zap_cursor_serialize(zap_cursor_t *zc);
371 * Advance the cursor to the attribute having the given key.
373 int zap_cursor_move_to_key(zap_cursor_t *zc, const char *name, matchtype_t mt);
376 * Initialize a zap cursor pointing to the position recorded by
377 * zap_cursor_serialize (in the "serialized" argument). You can also
378 * use a "serialized" argument of 0 to start at the beginning of the
379 * zapobj (ie. zap_cursor_init_serialized(..., 0) is equivalent to
380 * zap_cursor_init(...).)
382 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
383 uint64_t zapobj, uint64_t serialized);
386 #define ZAP_HISTOGRAM_SIZE 10
388 typedef struct zap_stats {
390 * Size of the pointer table (in number of entries).
391 * This is always a power of 2, or zero if it's a microzap.
392 * In general, it should be considerably greater than zs_num_leafs.
394 uint64_t zs_ptrtbl_len;
396 uint64_t zs_blocksize; /* size of zap blocks */
399 * The number of blocks used. Note that some blocks may be
400 * wasted because old ptrtbl's and large name/value blocks are
401 * not reused. (Although their space is reclaimed, we don't
402 * reuse those offsets in the object.)
404 uint64_t zs_num_blocks;
407 * Pointer table values from zap_ptrtbl in the zap_phys_t
409 uint64_t zs_ptrtbl_nextblk; /* next (larger) copy start block */
410 uint64_t zs_ptrtbl_blks_copied; /* number source blocks copied */
411 uint64_t zs_ptrtbl_zt_blk; /* starting block number */
412 uint64_t zs_ptrtbl_zt_numblks; /* number of blocks */
413 uint64_t zs_ptrtbl_zt_shift; /* bits to index it */
416 * Values of the other members of the zap_phys_t
418 uint64_t zs_block_type; /* ZBT_HEADER */
419 uint64_t zs_magic; /* ZAP_MAGIC */
420 uint64_t zs_num_leafs; /* The number of leaf blocks */
421 uint64_t zs_num_entries; /* The number of zap entries */
422 uint64_t zs_salt; /* salt to stir into hash function */
425 * Histograms. For all histograms, the last index
426 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
427 * than what can be represented. For example
428 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
429 * of leafs with more than 45 entries.
433 * zs_leafs_with_n_pointers[n] is the number of leafs with
434 * 2^n pointers to it.
436 uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
439 * zs_leafs_with_n_entries[n] is the number of leafs with
440 * [n*5, (n+1)*5) entries. In the current implementation, there
441 * can be at most 55 entries in any block, but there may be
442 * fewer if the name or value is large, or the block is not
445 uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
448 * zs_leafs_n_tenths_full[n] is the number of leafs whose
449 * fullness is in the range [n/10, (n+1)/10).
451 uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
454 * zs_entries_using_n_chunks[n] is the number of entries which
455 * consume n 24-byte chunks. (Note, large names/values only use
456 * one chunk, but contribute to zs_num_blocks_large.)
458 uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
461 * zs_buckets_with_n_entries[n] is the number of buckets (each
462 * leaf has 64 buckets) with n entries.
463 * zs_buckets_with_n_entries[1] should be very close to
466 uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
470 * Get statistics about a ZAP object. Note: you need to be aware of the
471 * internal implementation of the ZAP to correctly interpret some of the
472 * statistics. This interface shouldn't be relied on unless you really
473 * know what you're doing.
475 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
481 #endif /* _SYS_ZAP_H */